Oracle SQL比较同一表的行

Question

Table:

Date         shopId        amount
17-MAY-19      1             100
17-MAY-19      2             20
16-MAY-19      2             20
17-APR-19      1             50

I need to find those records of the same table according to the requirements:

Given a date eg sysdate, find the record with that date for each shopId ( the date is unique for each shopId) and also the pre-30day record.

Compare the amount of the 2 records to see if the absolute % diff is greater than 5 (I use 0.05 instead of % in the code below).

If both the 2 records exist and the % diff is matched, the record should be in the result set.

Do this for all shopId s and return the result set.

I am able to retrieve the records and then do the comparison in backend languages such as JAVA and PHP but I wonder if it is better to do it by SQL which I'm not familiar with.

select * 
from table t1 
inner join table t2 on t1.shopId = t2.shopId 
WHERE t1.ordertime = sysdate 
  and t2.ordertime = sysdate - 30 
  and abs( (t1.amount - t2.amount) / t2.amount > 0.05 ) 

Expected result should be:

Table:
shopId     Date1     amount1      Date2     amount2
  1      17-MAY-19    100       17-APR-19      50

Please help, thank you.

Answer 1

you should use trunc function on sysdate , because sysdate always contains the current date and time

with tab as(
  select to_date('17.05.2019','dd.mm.yyyy') as dat, 1 as shopid, 100 as amount from dual union all
  select to_date('17.05.2019','dd.mm.yyyy') as dat, 2 as shopid, 20 as amount from dual union all
  select to_date('16.05.2019','dd.mm.yyyy') as dat, 2 as shopid, 20 as amount from dual union all
  select to_date('17.04.2019','dd.mm.yyyy') as dat, 1 as shopid, 50 as amount from dual 

)

select * 
from tab t1 
join tab t2 on t1.shopid = t2.shopid 
WHERE t1.dat = trunc(sysdate)
  and t2.dat = trunc(sysdate - 30)
  and (t1.amount - t2.amount) / t2.amount > 0.05

Result

Answer 2

An alternative to joins would be to use the LAG analytic function, eg:

WITH your_table AS (SELECT to_date('17.05.2019','dd.mm.yyyy') dt, 1 shopid, 100 amount FROM dual UNION ALL
                    SELECT to_date('17.05.2019','dd.mm.yyyy') dt, 2 shopid, 20 amount FROM dual UNION ALL
                    SELECT to_date('16.05.2019','dd.mm.yyyy') dt, 2 shopid, 20 amount FROM dual UNION ALL
                    SELECT to_date('17.04.2019','dd.mm.yyyy') dt, 1 shopid, 50 amount FROM dual)
SELECT shopid,
       date1,
       amount1,
       date2,
       amount2
FROM   (SELECT shopid,
               dt date1,
               amount amount1,
               LAG(dt) OVER (PARTITION BY shopid ORDER BY dt) date2,
               LAG(amount) OVER (PARTITION BY shopid ORDER BY dt) amount2
        FROM   your_table
        WHERE  dt IN (TRUNC(SYSDATE), TRUNC(SYSDATE) - 30))
WHERE  amount1 > amount2 * 1.05;

    SHOPID DATE1          AMOUNT1 DATE2          AMOUNT2
---------- ----------- ---------- ----------- ----------
         1 17/05/2019         100 17/04/2019          50

Whether this is faster than the join version is up to you to test and decide.